Tangotiger Blog

Wednesday, November 16, 2016

What would have happened if the AL Cy Young used an instant run-off system?

?I explained the method several years back.  First one to get 15+ votes wins.  Let's do this:

Round 1:

14. Verlander

8. Porcello

5. Britton

3. Kluber

We knock out Kluber, and now we look at those 3 specific ballots, and see who those three voters put as #2.  They move to #1 with Kluber knocked out.  As it turns out, all 3 voters had Porcello #2.

Round 2:

14. Verlander

11. Porcello

5. Britton

We knock out Britton, and now we look at those 5 specific ballots, and see who those five voters put as #2.  They move to #1 with Britton knocked out.  As it turns out, all 5 voters had Porcello #2.

Final:

16. Porcello

14. Verlander

Monday, October 03, 2016

Cy Young Race, 2016, and the limits to modeling human behaviour

?So, this one was very tough to try to model.  I think it's likely that Jon Lester will take it, over Max Scherzer.  There is a slight advantage to Max in Wins (1), and a tiny advantage to Lester in Losses (2).  That's either in balance, or a very slight advantage to Max.  But it's pretty invisible.  In ERA, Lester is ahead by 11 runs.  When we draw lines, you have to draw lines.  So, you can draw the line that Max is ahead in Wins, or draw the line that Max is ahead in W-L/3.  In which case, you then have to decide how to balance Lester's 11 runs ahead to Max's slight lead in W.  Is the overwhelming lead in ERA enough that we don't need to look at secondary stats (IP and  K, of which Max would take over from Lester)?  Do we choose W-L/2, in which case Max is tied with Lester, and so, Lester takes it without looking at secondary stats?

Bill James has the New Season Score, which gives Scherzer 310 points to Lester's 302 points.  I have Max with 88 points to Lester's 85 points.  But as I've said, BEFORE we look at the points, we focus only on ERA and W (or W and some portion of Losses).  And I think Lester's overwhelming lead in ERA with the tiny disadvantage at best in W or W/L is enough that we don't need to rely on the points system to break the tie.

Hence, Lester for the W.

And the same thing repeats itself in the AL, but flipped the other way.  Porcello is 4 wins ahead of Kluber (with 5 fewer losses), while being behind him in ERA by 0.2 earned runs.  Which is a rounding error, which would therefore give the win to Porcello has on these two stats.  Similarly, Porcello is 6 wins ahead of Verlander (with 5 fewer losses), while behind Verlander in ERA by 2.5 earned runs.  Is it enough of an advantage for Verlander that we want to go to the secondary stats (of which Verlander has the advantage because of the strikeouts)?  It seems to me that it doesn't need to get there, that Porcello has built up enough of an advantage in W and L, and his ERA gap is just over the line as to not make a difference.

In terms of the New Season Score, Porcello is way ahead of not only Verlander and Kluber, but also Happ, who is ahead of these two guys.  But Porcello is behind Britton, 327 points to 309 points.  In terms of Cy Points, if we needed to go there, it's Verlander at 78, Porcello at 74, Kluber at 73.  Given that Porcello will either get 1st or 2nd place votes, while Britton can place anywhere on the ballot, or even not appear on a ballot, Britton could get more 1st place votes for Porcello, but that won't matter overall.

Hence, Porcello for the W.  

The tough part is the more down ballot picks, and I'll come up with my official prediction tonight.

Sunday, October 02, 2016

How will the NL Cy voters treat Johnny Cueto?

?There may be at least 11 SP in the NL that will receive at least 1 Cy Vote.  The voters are going to have a very tough time trying to figure out who will get to be on their ballots.  The top 3 contenders all  have a story to tell.  You have Lester and Scherzer leading the league in wins at 19.  Hendricks leads the league in ERA. Lester is 2nd in ERA, and Scherzer leads the league in K.  Based on those three stats alone, the top 3 contenders are these 3.  That leaves two spots on the ballot to choose from 8 other pitchers.

Cueto's story is that he's at 18 wins, putting him in 3rd place behind the big 2.  Fernandez (RIP) and Bumgarner are the only other ones with impressive K totals.  Syndergaard has impressive K rates and the better ERA.  And Kershaw has his own abbreviated story.

Guys at the bottom of the pack just won't get any love: Roark, Arrieta, Carlos Martinez.  For example, whatever Arrieta's got going for him, Cueto is a bit above him.  Roark is supplanted by Thor for the same reason, and Martinez brings up the rear with every other pitcher.

So, we can see 2 of the slots going to Bumgarner, Fernandez (RIP), Syndergaard, Kershaw based on individual preferences.  Cueto?  Well, I can see how he places third from this group of 5 on each ballot.  But with only 2 spots on the ballot, I can see Cueto getting almost no love at all.

This is how the readers see it as well:

https://twitter.com/tangotiger/status/782596425402224640

Cy Comps: 2016 Porcello/Verlander v 2012 Price/Verlander

?In 2012, Price edged out Verlander by 1 1st place vote.  That season, Price was 20-5 compared to Verlander's 17-8.  In my system, that's a 3 point advantage for Price.  Price had a 2.56 ERA compared to Verlander's 2.64.  In my system, that's a 2 point advantage for Price.  Because Price is able to squeak by on the two main categories, that's enough for me to have called it, and I predicted Price would get the Cy Young in 2012.  But boy was it close.  Given that there's hundreds of potential votes, of which they selected 28 (at the time), it could very well have gone the other way.  And that's because some voters don't follow the script, and they look at IP and K, among other things.  For IP, Verlander he gets 6 points for his 17 IP lead, and for K, Verlander gets 3 points for his 34 K lead.  So, that's 9 extra points for Verlander, compared to the 5 points for Price in the big 2.  But like I said, if you lead in the big 2, the extra points don't count.

In 2016, it's not just two pitchers, but six, that are in contention.  On the one side, you have Porcello and Happ, where their argument is mostly win-based.  Porcello is 22-4, 3.15, while Happ is 20-4, 3.18.  Porcello is 28 IP ahead, and 26 K ahead.  Because Porcello is ahead across the board, Happ will get relegated to lower-ballot status.  Porcello is the one that will get 1st place and 2nd place ballots, and maybe some third, while Happ will get his ballots 3rd through 5th, and maybe off the ballot for several voters.

On the other side we have the more traditional leaders: Verlander, Kluber, Sale.  Similar to Verlander in 2012, their ERA is close to Porcello (3.10 to 3.21).  I think the closeness of the ERA, not to mention close to the league leaders Tanaka at 3.07 and Sanchez at 3.06 (albeit at under 200 IP for both) will simply mean that the ERA is going to be a wash.  They are all effectively tied.  Which leaves the 1st place voting based just on W, which will go to Porcello.

But for 2nd place, with Porcello out of the mix, things get more interesting.  Will the same thought process take place, and then just hand it to Happ?  Or, having already given it to Porcello, there's going to be more balance to the ballot.  And that's where the Cy points system comes into place.  And that calls for Verlander and Kluber to get 2nd and 3rd (though we'll see tonight if Verlander has a bad game or not).  

Then 4th and 5th goes back to Happ v Sale, and that looks like 2012 Price v Verlander. Happ is 3 wins ahead of Sale, and his ERA is 3.18 to 3.21.  Both squeakers of a win, just like Price/Verlander.  It doesn't stop there.  Sale has a 27 IP lead, similar to Verlander/Price, and 64 K lead.  Not to mention the CG.  It'll be close between the two.

Then we have Britton, who will get votes from 1st through 5th, and then off ballot.  If we try to come up with how the voting will go by 1st through 5th, here's one that is off the top of my head:

Pts 1 2 3 4 5 Off Pitcher
161 14 15 1 0 0 0 Porcello
103 7 4 8 5 4 2 Verlander
89 5 4 6 7 6 2 Kluber
65 4 4 4 3 3 12 Britton
42 0 3 4 6 6 11 Sale
38 0 2 5 5 5 13 Happ
10 0 0 1 2 3 24 Tanaka
10 0 0 1 2 3 24 Sanchez

Is this how the voters are going to think?  Take Porcello out of the mix, and it might be Happ who would otherwise win it.  But, since Porcello is getting all the W-L votes, Happ won't get similar consideration.  Britton is going to get sprinkled in all over, but as you can see, with so many viable candidates, and only 5 spots, there's going to be plenty of pitchers left off ballots.  And I haven't even brought up Hamels and Quintana either.

And remember, not all voters are voting.  The pool is  made up of hundreds of potential voters, of which 30 are selected for the AL Cy.  So, it's very easy to see how you can randomly choose 30, and from those 30, it's Britton that ends up in 2nd place, or Britton that ends up in 7th.  With not  much to distinguish them, it becomes a tough job to predict.

Anyway, after tonight I'll come up with my final forecasts.

Wednesday, September 21, 2016

Cy Young race - Lester and Porcello now in the lead

?As of today:

Head-to-head: BBWAA would vote Lester over Hendricks, Lester over Bumgarner, Lester over Scherzer.

Head-to-head: BBWAA would vote Scherzer over Hendricks, Hendricks over Bumgarner, Bumgarner over Scherzer.  That's a conundrum, as we have S > H > B > S.  

Head-to-head: BBWAA would vote Porcello over Sale, Porcello over Kluber.

Head-to-head: BBWAA vote tie Sale/Kluber.

That really just leaves Britton.  Realistically, Porcello would only get 1st, 2nd, 3rd place votes.  Britton can appear 1st through five, and even not on any ballots.  Britton might end up 2nd overall, and maybe 4th overall.  But not 1st.

Sunday, September 18, 2016

Trying to model the BBWAA voting behaviour

?So, this year is proving to be quite challenging.  I'm hoping that something is going to happen in the next 2 weeks to break the logjam.  The system I had was straight forward: figure out if anyone leads in both W and ERA.  If you find that guy, put him on the list.  If you don't find someone like that, find the guy who has the most Cy Young points and put him on the list.  In this case, that would be Max Scherzer.

Now that he's out of the pool, we start over.  Of those remaining, does someone lead in W and ERA?  No.  Then go to the Cy Points, which is Bumgarner.  He's now 2nd. 

And we keep going.  Lester comes in third, and Hendricks 4th.  

The problem we have is Kyle Hendricks.  Because he leads the league in ERA and with only 15 wins, the ERA/Wins check has no hope at being applied to the top of the ticket.  Now, let's see what happens if Kyle Hendricks has a couple of rough outings so that he's no longer in the ERA lead, and it is instead Lester that leads the league in ERA.  Well, suddenly we have Lester that leads the league in ERA and is tied for the league-lead in Wins.  This automatically puts him at the top of the list.

Does Hendrick's current place atop the ERA leaderboard impact how the voter sees Lester?  I have no idea!  We can make the argument that the way the voter behaves is this way:

1. Put everyone in order of Cy Points.
2. For each player, look at everyone above him.  If someone above him has both fewer W and a worse ERA, put that guy above him.

In this case, we start with Scherzer, Bumgarner, Lester, Hendricks in our list.  Then, we see that Lester has both more W and better ERA than Bumgarner, so Lester bumps ahead of Bumgarner.  And Hendricks bumps ahead of Bumgarner as well.  We end up with Scherzer, Lester, Hendricks, Bumgarner.

It seems that both approaches are the same, but this little nuance throws everything off.  So, I think I have to retool the way I have my program setup to handle this particular approach.  Is this approach going to better model voter behaviour?  Like I said, I have no idea.  You can make the case that Lester needs to lead all the pitchers in W, not just be tied for the league-lead.  Does Lester and Scherzer both having 17 W automatically allow Lester's 2.40 ERA to override Scherzer's 2.78 ERA?  And we don't bother with Scherzer's 80K lead?  I don't know, it's a tough call.  But if Lester had 18 W to Max's 17, then I'd say: yes, it does.

Like I said, let's see how things shape up at the end.  Right now, it's a 4-man race, with Thor and Cueto still having a chance to impact the race if they dominate in their remaining starts.

Tuesday, August 16, 2016

Relief pitchers and Cy Young

?Last week we were talking about the relievers and the Cy Young.  I forwarded that discussion to Bill James, and in true Jamesian style, he went and did an all-out study on it.  He used a new version of Season Scores which, by the looks of it, could even be a replacement to his Cy Young predictor, or at least a sibling of it.

Friday, May 13, 2016

Cy Young odds, 2016

Taken with a grain of salt...

Wednesday, April 27, 2016

On Pace to win the 2016 Cy Young

?Like last week, it's still Kershaw and Sale, as likely 50/50 each.

Arrieta has put a bit of distance from the other contenders, and Syndergaard has joined the fray with Scherzer, while Bumgarner, Strasburg and Cueto are ready to pounce.  Otherwise, there's only Lester and Jose Fernandez who even have a shot at contending for the #2 spot.

Over in the AL, the race for #2 has the usual suspects of Price, Keuchel, Kluber, Felix, plus a pack of another dozen pitchers.  

Thursday, April 21, 2016

On Pace to win the 2016 Cy Young

A lot of foolish people think that "on pace" means "pro-rated". That's why you don't listen to foolish people. It's like with the elections: you don't just pro-rate known votes and declare an "on pace" winner. You take what is already banked and you forecast the remaining time, and combined, you get your "pace".

Based on pre-season Steamer forecasts, Kershaw was on pace to win the 2016 Cy Young, and he still is. The next 4 guys that were on pace to win, based on pre-season, were Arrieta, Scherzer, Bumgarner, and Greinke. Greinke as of today is almost already eliminated from that pace. Instead, Syndergaard, Strasburg are now on pace to be contenders.

Over in the AL, Sale is already on pace to run away with the Cy Young, as likely as Kershaw is to win the NL. After that, Price has dropped from strong contender to part of a crew that includes Keuchel, and to a lesser extent Carrasco and Felix.

Friday, April 08, 2016

Could Quadratic Voting have effected the Cy Young?

?Suppose the voter is giving 100 points to distribute.  But if he gives multiple points to a pitcher, it counts squared.  Something like this:

7 points Kershaw (49 credits)

5 points Greinke (25 credits)

4 points Arrieta (16 credits)

That's a total of 90 credits.  He can still hand out another 10 credits.  He can list 10 pitchers at one point each.  Or he can give out 3 points to Gerrit Cole (9 credits) and 1 point to Bumgarner (1 credit).  That's 100.

Compare this PARTICULAR allocation, of 7-5-4-3-1 to the STANDARD allocation of 7-4-3-2-1.  If you really wanted to keep Kershaw/Greinke/Arrieta closer together, you can.  

Thursday, March 31, 2016

Favorites for 2016 Cy Young

?Neil brings up Thor, so, let's take a look at Steamer and see who are the favorites.

In the NL, Kershaw is easily the favorite, and I'd estimate 50/50 shot.  After him, it's Scherzer, Bumgarner, Arrieta, Greinke.  I'll guess that 2-3 of those 4 will finish in the top 5. After that, it's "Cy breakout" potential from Jose Fernandez and Strasburg.  One of Lester, Harvey, deGrom, Cueto, Syndergaard and Gerrit Cole should be able to finish in the top 5.

Moving over to the AL, Sale is the favorite with Price close behind.  Keuchel and Kluber are in the mix.  After that, it's Carrasco, Archer, and Felix.  Three or four of the top 5 will come from this group of 7.

Friday, January 22, 2016

How to make one vote count for 10

?I have an article over at BPro that discusses how the IBA results would have changed with a first-past-the-post system, and I use the IBA votes in an instant-runoff method.

To get a bit more technical on the various voting impacts: in this case, it was a clear two-man race, Trout v Donaldson.  The vast majority of the voters had these two players somewhere at 1-2.  The rest of the players on the ballot were simply noise.  And in "ranked" voting (whatever the official term is), you give 10 points for whoever you ranked 1st and 9 points for second, and so on (though in BBWAA MVP, it's 14, then 9).  If your ballot shows Trout-Donaldson, you are giving 10 to Trout and 9 to Donaldson. Accepting the rest of the players as noise, this is EXACTLY the same thing as giving 1 to Trout and 0 to Donaldson.  Indeed, listing Trout 10th on your ballot, and leaving Donaldson off altogether is IDENTICAL in impact to listing these players 1-2.  Therefore, you give 1 more point to Trout than Donaldson. 

If you want to give Trout TWO more points than Donaldson, you list them 1-3 or 2-4 or 3-5... or 8-10 or 9 and leave the other guy off the ballot.  If you want to give Trout THREE more points than Donaldson, you place Trout 8th and leave Donaldson off the ballot.  Four more points for Trout?  List Trout 7th and leave Donaldson off.  And on and on we go.  List Trout 1st, leave Donaldson off the ballot altogether, and you are giving Trout 10 points.  And your vote now counts for ten times as much as an honest vote.

Indeed, one guy placing Trout 1st and leaving Donaldson off altogether will exactly cancel out the ten voters placing Donaldson first and Trout 2nd.

I'll come back later in the morning to show the breakdown as to how the BPro readers voted on this.

Thursday, January 07, 2016

HOF 2016 results

?Courtesy of a tireless researcher over at Bill James' site.  If it's blocked, let me know.

Thursday, December 31, 2015

Simple HOF prediction system

?1. Start with Ryan Thibs data.

2. Figure out the gap between non-public and public votes in 2015 and apply to 2016.  For new guys, make a reasonable estimate.  I set Junior at -4%, and the others at 0%.

3. Apply that gap to 2016 data of public ballots.

4. Figure the weighted average.

Here are the results.  You can of course spend more time trying to refine this, look at past history on these players, and on historical data.  The below is your starting point, the Marcel-version.

97% Ken Griffey Jr.
81% Mike Piazza
75% Jeff Bagwell
70% Tim Raines
65% Trevor Hoffman
50% Curt Schilling
49% Alan Trammell
48% Edgar Martinez
46% Roger Clemens
46% Mike Mussina
46% Barry Bonds
39% Lee Smith
20% Fred McGriff
18% Larry Walker
16% Mark McGwire
15% Jeff Kent
11% Sammy Sosa
10% Gary Sheffield
9% Billy Wagner
6% Nomar Garciaparra
2% Jim Edmonds
1% Garret Anderson

Wednesday, December 30, 2015

Fighting for Mike Mussina into the HOF

?The hidden story of the 2016 ballot is the fight to get Mike Mussina the support for the HOF.  Thanks to the tireless work of Ryan Thibs, we learn that Mussina has some powerful support.

Last year, the BBWAA who made their ballots available to the public had Mussina with 30% yes votes.  For those who did not release their ballots, the support was only 16%.  That gap was third largest in 2015, beat barely by Curt Schilling (who the writers seem to hate), and... Pedro?  Yes, apparently those who didn't vote for Pedro were too ashamed to admit it.

Anyway, ,this year, Mussina is showing the largest jump of all players between 2015 and 2016.  Of the public ballots anyway.  The real interesting thing will be to see if the private ballots are going to also jump, or, for some reason, the gap between public and private will widen even further than the already wide gap.

Tuesday, December 22, 2015

If Fans filled out their HOF ballots 10-deep, what would the results look like?

?Using the results from the head-to-head matchups from the Straight Arrow readers, I can come up with the odds of any one player beating any other player.  For example, on a one-man only ballot, Bonds would get 58% of the votes, Junior would get 11%, Clemens would get 9%, Raines+Bagwell+Piazza would get 9%, and everyone else would get 14%.  And no one gets voted in.  On a two-man ballot, Bonds would get 88%, and Junior would get 27%.   You'd need to get to a 5-man-must ballot for Junior to get elected: Bonds appears on every ballots, while Junior appears on 79% of the ballots.  With a 6-man ballot, Clemens makes it.

When every ballot must list 10 names, this is what happens:

100%    Barry Bonds

100%    Ken Griffey

100%    Roger Clemens

81%    Jeff Bagwell

81%    Tim Raines

80%    Mike Piazza

58%    Curt Schilling

55%    Mike Mussina

48%    Edgar Martinez

43%    Alan Trammell

41%    Larry Walker

39%    Mark McGwire

31%    Gary Sheffield

28%    Sammy Sosa

25%    Jim Edmonds

21%    Jeff Kent

21%    Fred McGriff

16%    Trevor Hoffman

12%    Nomar Garciaparra

11%    Billy Wagner

8%    Lee Smith

Six players get elected.  Now, this is a good argument AGAINST filling out your ballot.  But, since we know there are plenty of voters that don't fill out their ballot, then this becomes part of gamesmanship, to undo the damage of others who have light ballots.  According to the fans, Bonds is such an overwhelming candidate, that you need to list 5 players to elect 2, and 6 to elect 3.  That really ends up becoming the minimum of any ballot, to list six candidates, to counteract part of the vote-splitting effect.

Friday, December 11, 2015

Hall of Fame, Head-to-Head

This is the fourth year I’ve run this, and I’ve been running this kind of project since 2010.  You will be given two players from the current Hall-of-Fame ballot, and you choose the player that you think had the more outstanding career.  That’s it.  You try again and again, and I compile all the results.  Have fun!

http://www.tangotiger.net/hall/

Piazza-not-Bagwell v Bagwell-not-Piazza voters

?Terrific piece by Sam, as he goes inside the minds of the BBWAA voters.  Love this stuff.

Sunday, December 06, 2015

Who will get the Johnson, Pedro, Smoltz, Biggio votes?

?As we learned last year, there's a relationship between the quality of players removed from a ballot, and how those votes get redistributed.  There are 3.65 votes that went to players who were voted in, or dropped off in the last ballot.  About half of that is going to be taken by the new guys (Junior, Hoffman, Edmonds, Wagner, et al).  So, there's about 1.8 or so votes per ballot that are "lost".  Based on the above research, this means 1 vote per ballot will go to the returnees.

There's 4.75 votes that went to players returning onto the ballot, meaning that they will now get 5.75 votes, an increase of 20%, or an increase of around .06 votes per player.

On top of that however, we're looking at "the purge" (voters no longer eligible), who we have learned were predisposed to have fewer players per ballot than the remaining votes.  That will give each player an extra .03 votes per player, say another 0.6 votes per ballot.

All in all, we're looking at returnees jumping by around .10 votes per ballot each, with a total of around 8.2 players per ballot.